Developers
Building on the Sume API: keys and hosts, jobs and runs, polling, webhooks, idempotency, errors and rate limits, the SDK, the CLI, and MCP auth.
MCP OAuth without DCR: client ID metadata documents and issuer checks
MCP 2026-07-28 deprecates dynamic client registration for Client ID Metadata Documents and requires clients to validate iss. What it means for Sume.
Cursor self-hosted machines can't take Sume webhooks on a private URL
Sume rejects localhost, private-network and non-HTTPS webhook URLs. An agent on a self-hosted machine should poll status_url or use a public HTTPS receiver.
Sign MCP requestState for paid render approvals: user, TTL, digest
MCP says requestState is attacker-controlled. For a paid render approval, bind it to the user, an expiry and an argument digest with an HMAC. Python included.
Python 3.10 is end of life: a stdlib Sume webhook verifier
Python 3.10 has reached end of life. A standard-library verifier for Sume's signed webhooks that refuses an empty secret and accepts rotated signatures.
Python MCP SDK 2.3 subscriptions=False: a Sume wrapper that pulls
MCPServer(subscriptions=False) turns off push subscriptions in the Python SDK. A wrapper over Sume jobs can do without them by waiting in bounded slices.
Demand Gen copy limits: 40-character headlines, one at 30 or fewer
Demand Gen allows 40-character headlines (one must be 30 or fewer), 90-character descriptions and 10-60 second videos. A checker script plus the Sume lengths.
Push or poll for a finished render: listen, webhook or jobs_wait
MCP 2026-07-28 adds subscriptions/listen. For a render that takes minutes, compare a listen stream, a signed webhook and jobs_wait, with a Python verifier.
OpenAI Agents API sandbox: keep the Sume API key out of it
The OpenAI Agents API beta adds a sandbox and hosted-browser computer use. Where a Sume API key can live when an agent runs there, and what to hand it instead.
Grok Imagine video API: 15 s, 5 references, request-ID polling
xAI's video guide for grok-imagine-video-1.5 lists up to 15 seconds, up to 5 reference images and async polling by request ID. The same loop on Sume jobs.
Shopify rejects file names ending in thumb, icon or large
Shopify file uploads reject names ending in pico, icon, thumb, testing, small, compact, medium, large or grande. A Python rename step for batch outputs.
Pause between narration lines: TTS pause markers or audio concat?
Deepgram Flux TTS allows pause markers of 500 to 3000 ms, max 8 per request. Sume's timeline audio concat joins lines with no gap. Where to put the pause.
Shopify image limits: 20 MB, 25 megapixels vs Sume image outputs
Shopify accepts product images up to 20 MB and 25 megapixels in JPEG, PNG, WEBP, HEIC or GIF. How that lines up with Sume image model sizes and formats.
Design a render tool for a stateless MCP server: job ids as arguments
MCP 2026-07-28 removes protocol sessions. A render tool stays correct if its state lives in a job id the client passes back, as Sume jobs do.
How long AI video vendors keep your file: Veo 2 days, Higgsfield 7+
Veo keeps videos two days, Higgsfield files at least seven, Sora Batch outputs were kept 24 hours. Retention facts and a download-on-complete script.
Recraft V4.1 Flash: median 1.3 s, p95 1.8 s. Set timeouts from p95
Recraft quotes a median of about 1.3 seconds and a p95 of 1.8 seconds for V4.1 Flash. How to turn latency claims into timeouts and polling for image APIs.
VS Code 1.140 shared MCP config files: what goes in the Sume entry
VS Code 1.140 lets MCP servers live in portable config files shared across Copilot tools. For Sume the entry is one URL, and no key belongs in the file.
Text to Dialogue continuity: 100-character context, 3 request IDs
ElevenLabs' Sept 28 changelog adds previous_text and future_text (100 chars max) and request-ID chaining (3 max). A Python limit check.
Keep your Sora-style create_video() call: map it onto Sume
Sora's seconds, size and input_reference become duration, resolution plus aspect_ratio, and a first frame. Here is that map as a Python wrapper over Sume.
Claude Code 2.1.285 lists WebSocket MCP servers; Sume uses HTTP
Claude Code 2.1.285 shows WebSocket MCP servers in claude mcp list. Sume's hosted MCP is a remote HTTP server, added with --transport http.
C2PA 2.2: file types that can carry credentials vs Sume outputs
C2PA 2.2 manifests can be embedded in JPEG, PNG, WebP, SVG, MP4, MOV and more. How that list lines up with the formats Sume image and video jobs return.
Claude Code 2.1.289 plugin loading fix: recheck Sume MCP after upgrade
Claude Code 2.1.289 fixed plugin loading after an upgrade. A four-call read-only check confirms the Sume MCP connection and scopes before you run a paid job.
Cursor Security Review bot on a Sume webhook handler: what to find
Cursor added a Security Review bot on Sep 23. A webhook handler for Sume should pass seven checks: raw body, timestamp window, rotation, empty secret and more.
C2PA 2.2 in plain terms: manifests, hard and soft bindings
C2PA 2.2 describes signed manifests with hash-based hard bindings and fingerprint or watermark soft bindings. What that implies after a re-encode or trim.
What a media MCP server should declare at server/discover
MCP 2026-07-28 adds a required server/discover call. A media server has more to say than versions: async jobs, wait limits, scopes. Where Sume documents each.
Gemini CLI v0.63 plan execution in CI: gate paid Sume calls first
Gemini CLI preview v0.63.0 adds autonomous plan execution in non-interactive mode. Before unattended runs, gate Sume paid tools with dry_run and max_spend_usd.
Workflow DevKit 5: 12 MiB frames, pass Sume URLs not video
Workflow DevKit 5.0.1 splits WebSocket frames over 12 MiB by default (16 MiB at most). Pass Sume artifact URLs between steps instead of video bytes.
A "use step" function that submits a Sume job once
Put the Sume submit call inside one "use step" function and derive the Idempotency-Key from a stable run key, so a retried step returns the original job.
Expired MCP session after 30 idle minutes: resume by job id
MCP Python SDK v2.2.0 closes idle Streamable HTTP sessions after 30 minutes by default. Keep the Sume job_id and resume the render with jobs_wait.
LangGraph custom image: keep SUME_API_KEY out of it
langgraph-cli 0.4.32 adds an --image-uri flag for self-hosted custom containers. Inject SUME_API_KEY as a runtime env var; never bake it into the image.
Headless Gemini CLI and MCP auth: use a Sume API key
A headless Gemini CLI run cannot finish a browser consent. Connect to hosted Sume MCP with an API key header instead, and keep paid calls bounded.
Railway webhook custom headers vs Sume's signed headers
Railway project webhooks now accept custom headers. Sume signs its callbacks with x-sume-webhook-timestamp and x-sume-webhook-signature; verify them in Python.
Container snapshot restore: resume a Sume job from its job_id
Cloudflare Containers can snapshot a running container. Save the Sume job_id in it, and after a restore read the job's status_url instead of resubmitting.
OpenAI Agents API hosted sandbox: which Sume hosts to allow
OpenAI's Agents API is in public beta with hosted or connected sandboxes. Which Sume hosts to allow, how to pass the MCP URL, and why the key stays in a secret.
A Go client for Sume from OpenAPI, with Retry-After
The docs list only a TypeScript SDK. Generate a Go client from the live OpenAPI schema, send x-api-key, and back off on 429 with retry-after.
Voice agent hand-off: ask for a clip, get an async Sume job
A live voice agent should not wait on a video render. Hand the request to an async Sume job, speak the job id back, and deliver the clip by poll or webhook.
Temporal Paygo starter: submit and poll a Sume job
Temporal's Paygo plan has a $0 monthly minimum. A first workflow can submit a Sume job in one Activity and poll its status in a second. Python code included.
OpenAI Python 3.23 turn artifacts: hand over a Sume render
OpenAI Python SDK 3.23.0 adds file staging and turn artifact downloads for agents. Sume job webhooks give you a public media.sume.com URL to hand over.
Workflow DevKit 5 fail-fast refusals: which Sume errors to retry
Workflow DevKit 5.0.1 now fails fast on dynamic start() refusals. Apply the same rule to Sume: do not retry a 400 or 402, retry 429 and 503 with the same key.
Omni Flash prompts in English, captions in your language
Google says Omni fully supports English; other languages are unevaluated. Prompt in English, then add captions in your language with Sume's captions API.
IT admin checklist: approving the Sume MCP connector
The MCP 2026-07-28 post proposes mandatory OAuth and OIDC discovery support. This checklist covers Sume's hosted MCP OAuth, scopes and keys.
Copilot CLI dynamic workflows: let them call the Sume CLI
GitHub added dynamic workflows to Copilot CLI on Oct 1, 2026. Give it the Sume CLI skill pack and the hosted MCP endpoint, and keep paid calls behind a gate.
Pydantic AI slot leak vs Sume queue_full: tell them apart
Pydantic AI v2.53.0 fixed a streamed-request concurrency slot leak. A client limiter is not Sume's workspace queue_full 429, and each needs its own handling.
Explicit key vs saved login: auth order in the Sume CLI
The Sume docs state no precedence between a saved login and an env key. Here is what they do say about saved login, env keys, auth mode and the two-header 401.
Wan 3.0 at 30 fps and 2 to 30 s: the Sume model id
Model Studio lists Wan 3.0 at 480P to 1080P, 2 to 30 s, 30 fps. On Sume the id is wan-3.0, 2 to 30 s; first and last frames go in frame_images.
Railway Sandboxes: run the Sume CLI with an env-var key
Inside a short-lived sandbox, install the Sume CLI with the hosted installer, pass SUME_API_KEY as an environment variable, and never print it.
Three Sume throttle signals: rate_limited, queue_full, 503
OpenAI now splits 429 (traffic rising too fast) from 503 (overload). Sume has three: 429 rate_limited, 429 queue_full and 503 provider_capacity_exceeded.
MCP tool ran twice: idempotency_key saves your Sume render
Claude Code fixed MCP tool calls that sometimes ran twice on large results. Sume's paid MCP tools require an idempotency_key so a repeat does not bill twice.
OS credential store vs sume login: where the key lives
Inngest v1.45.0 stores CLI OAuth credentials in the OS credential store. The Sume CLI stores its login key in ~/.sume-com/config.json; use env keys in CI.